iiinova — where your digital self keeps evolving

JohnMark Taylor @johnmarktaylor.bsky.social · 04/08/2026

Excited to be working as an @AnthropicAI Safety Fellow in the Bay Area from August-November, focusing on mechanistic interpretability. Drop me a line if you want to meet up or talk shop!

050

Reposted by JohnMark Taylor

Eivinas Butkus @eivinas.bsky.social · 21/07/2026

Do world models emerge in neural networks trained for prediction? Generative Adversarial Collaboration (GAC) at #CCN2026

1275

JohnMark Taylor @johnmarktaylor.bsky.social · 11/05/2026

Does anyone have a favorite platform for live Q&A during a talk session? Ideally would have: 1. Unlimited participants 2. Easy export, e.g. to a txt file 3. Easy sorting based on popularity or recency

010

JohnMark Taylor @johnmarktaylor.bsky.social · 04/05/2026

(Postscript: as much as we've learned with our multimillion dollar MRI scanners, how much more is there to learn about the brain from doing just the right simple experiment? Food for thought.)

020

JohnMark Taylor @johnmarktaylor.bsky.social · 04/05/2026

Anyways, you can try it and introspect for yourself. Let's give our field some credit (no physics envy required): a theory made bold and counterintuitive predictions that were subsequently borne out. And Treisman did it with nothing but old-timey tech and brilliant thinking. 7/7

110

JohnMark Taylor @johnmarktaylor.bsky.social · 04/05/2026

Later work has questioned the existence of illusory conjunctions, suggesting they may be artifacts and not genuine percepts. This raises an interesting philosophical puzzle: if people confidently report seeing them, why doesn't that settle the issue? 6/ psycnet.apa.org/record/1999-...

psycnet.apa.org

APA PsycNet

100

JohnMark Taylor @johnmarktaylor.bsky.social · 04/05/2026

At least for me, seeing the illusory conjunctions required a weird half-focused headspace—if I'm paying attention *too* closely to the letters I don't see them (exactly as her theory predicts!), but once I start zoning out they start popping up. 5/

100

JohnMark Taylor @johnmarktaylor.bsky.social · 04/05/2026

If you're like the original participants, you might even see the numbers as colored sometimes (they're black every trial—promise). 4/

100

JohnMark Taylor @johnmarktaylor.bsky.social · 04/05/2026

Finally, check your answers against the correct answer shown a few seconds later. Some trials, you may find that you saw "illusory conjunctions" that weren't there—for example, a green "N" and pink "T" when in fact a pink "N" and green "T" were present. 3/

100

JohnMark Taylor @johnmarktaylor.bsky.social · 04/05/2026

Out of curiosity I recreated her original experiment. Here it is (full-screen helps): each trial you'll briefly see a row of three colored letters, surrounded by two numbers. Try to report the numbers, then as many colors and letters as you can. 2/ www.youtube.com/watch?v=L3pr...

youtube.com

Illusory Conjunctions Demonstration

YouTube video by JohnMark Taylor

100

JohnMark Taylor @johnmarktaylor.bsky.social · 04/05/2026

My favorite vision science success story: based on her theory Anne Treisman boldly posited a weird new kind of conscious experience, where we see features (e.g., colors & shapes) combined the wrong way. Then she just went and created this experience in people's brains. 1/

151

Reposted by JohnMark Taylor

Seda Karakose-Akbiyik @sedaakbiyik.bsky.social · 30/04/2026

If there’s a concept in cognitive science everyone knows, it’s the binding problem. But being well known doesn’t mean there’s agreement on what it is, or that it’s been solved. Excited to be co-organizing a symposium at @vssmtg.bsky.social reimagining it for the 21st-century. Come join us!

061

JohnMark Taylor @johnmarktaylor.bsky.social · 09/04/2026

We'd like to thank our wonderful panelists for their help in planning this event across the last few years, some of whom are traveling a long way to join us, and to the @vssmtg.bsky.social staff for all their logistical support. We look forward to hearing your thoughts!

000

JohnMark Taylor @johnmarktaylor.bsky.social · 09/04/2026

We'd love to involve the broader VSS community in this event, starting now. We genuinely want to know where you stand on this topic, so please feel free to give us your open-ended hot takes here (the spiciest takes may be featured in the intro): docs.google.com/forms/d/e/1F...

docs.google.com

Binding Problem Hot Takes

112

JohnMark Taylor @johnmarktaylor.bsky.social · 09/04/2026

We think this is a perfect time to take stock on this topic—recent findings across diverse subfields invite us to reexamine nearly every facet of the "textbook" binding problem, and the DNN revolution has given us new tools and "model organisms" for exploring it afresh.

100

JohnMark Taylor @johnmarktaylor.bsky.social · 09/04/2026

If you're like us you might've wondered at some point: whatever happened with the binding problem? Is it alive and kicking, solved, ill-posed? We're delighted to have the chance to hash this out from all interdisciplinary angles (behavior, neuro, AI) at this year's VSS.

100

JohnMark Taylor @johnmarktaylor.bsky.social · 09/04/2026

🧵Excited to announce— "Reimagining the binding problem(s) for the 21st century": A VSS Symposium St. Pete Beach @vssmtg.bsky.social May 15th, 10:30am Presenters: Peter Tse, JohnMark Taylor, Seda Karakose-Akbiyik, Ana Chica, Anne Sereno, & Jake Quilty-Dunn visionsciences.org/symposia/?sy...

1144

JohnMark Taylor @johnmarktaylor.bsky.social · 18/04/2025

now on Twitter an AI dog automatically pops up and spouts regime propaganda slop at you when you express heretical opinions

030

Reposted by JohnMark Taylor

Barack Obama @barackobama.bsky.social · 15/04/2025

Harvard has set an example for other higher-ed institutions - rejecting an unlawful and ham-handed attempt to stifle academic freedom, while taking steps to make sure students can benefit from an environment of intellectual inquiry, rigorous debate and mutual respect. Let’s hope others follow suit.

15568916518162

JohnMark Taylor @johnmarktaylor.bsky.social · 15/04/2025

Excellent primer. Can be hard to pack in software engineering training amidst everything else in grad school but a little bit goes a long way

082

JohnMark Taylor @johnmarktaylor.bsky.social · 27/03/2025

gotta believe the top salsa dancers would score through the roof on spatial intelligence / shape rotation tests

010

Reposted by JohnMark Taylor

👻 Spooky Elle 🎃🧙🏼‍♀️🍄 @hecatestorch.bsky.social · 21/03/2025

Just found out NIH has officially canceled our grant, a longitudinal study of Alzheimer's disease in Black Americans. I cannot even put into words how angry I am. The truth is they are canceling it because it has Black in the title. That's it, there is no other reason to do this.

501825784

JohnMark Taylor @johnmarktaylor.bsky.social · 28/01/2025

I searched for “clave rhythm” and this came up

000

JohnMark Taylor @johnmarktaylor.bsky.social · 28/01/2025

Geometry of Musical Rhythm is a lovely book

000

JohnMark Taylor @johnmarktaylor.bsky.social · 27/12/2024

Thank you! Hope it helps you out and welcome any feedback you’ve got

020

JohnMark Taylor @johnmarktaylor.bsky.social · 27/12/2024

Great suggestion, that’s the biggest item on the to-do list and actually shouldn’t be too difficult with the existing infrastructure—just want to make the interface user-friendly and flexible so there are some design choices to figure out

030

JohnMark Taylor @johnmarktaylor.bsky.social · 27/12/2024

Deep neural networks are complex, but looking inside them ought to be simple. Check out TorchLens, a package that can visualize any PyTorch network and extract all activations and metadata in just one line of code: github.com/johnmarktayl... www.nature.com/articles/s41...

25414